Anda di halaman 1dari 12

Research Article

Articulated Arm Coordinate Measuring Machine Calibration by


Laser Tracker Multilateration
Jorge Santolaria, Ana C. Majarena, David Samper, Agustn Brau, and Jess Velzquez
Departamento de Ingeniera de Dise no y Fabricaci on, Edifcio Torres Quevedo, EINA, Universidad de Zaragoza, 50018 Zaragoza, Spain
Correspondence should be addressed to Jorge Santolaria; jsmazo@unizar.es
Received 27 August 2013; Accepted 3 November 2013; Published 29 January 2014
Academic Editors: G. Huang and D. Veeger
Copyright 2014 Jorge Santolaria et al. Tis is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
A new procedure for the calibration of an articulated arm coordinate measuring machine (AACMM) is presented in this paper.
First, a self-calibration algorithm of four laser trackers (LTs) is developed. Te spatial localization of a retrorefector target, placed
in diferent positions within the workspace, is determined by means of a geometric multilateration system constructed from the
four LTs. Next, a nonlinear optimization algorithm for the identifcation procedure of the AACMM is explained. An objective
function based on Euclidean distances and standard deviations is developed. Tis function is obtained from the captured nominal
data (given by the LTs used as a gauge instrument) and the data obtained by the AACMM and compares the measured and
calculated coordinates of the target to obtain the identifed model parameters that minimize this diference. Finally, results show
that the procedure presented, using the measurements of the LTs as a gauge instrument, is very efective by improving the AACMM
precision.
1. Introduction
In recent years, there has been an increasing interest in
AACMMs because of their advantages in terms of accuracy,
portability and suitability for inspection and quality control
tasks in machining tool processes and in the automotive and
aerospace industry [1].
Nevertheless, few researches have focused on the cali-
bration of these mechanisms. Moreover, there is an absence
of standards on verifcation and calibration procedures. For
that reason, AACMM manufacturers have developed its own
evaluation procedures. Tese evaluation methods are based
on the three main standards for performance evaluation in
current CMMs, UNE-EN ISO 10360, ASME B89.4.1, and
VDI/VDE 2617 and are still carried out today to compare and
evaluate the accuracy of an arm from the point of view of
the CMMs. In [1], the author presented a procedure to check
the performance of coordinate measuring arms by calculating
the distances between the centers of diferent spheres. Te
results obtained were compared with the application of the
ANSI/ASME B89 volumetric performance test showing good
agreement between the two approaches and a cost reduction.
In [2], Shimojima et al. presented a new method to estimate
the uncertainty of a measuring arm using a tridimensional
gauge. Tis method consists of a fat plate with 9 spheres fxed
at three diferent heights with respect to the metallic surface
of the plate. Ten the spheres centers are measured with
the measuring arm at diferent locations and orientations,
and distances between spheres centers are compared to the
nominal distances to evaluate the measuring performance of
the arm. Other works have been found in the literature whose
main goal is also to evaluate the performance of measuring
arms [35].
However, the AACMM presents diferent characteristics,
and diferent verifcation procedures are therefore required.
A point clearly defnes a position of the three machine
axes for a CMM. Nevertheless, the possible positions of the
AACMM elements to achieve a fxed point defned in the
measurement volume are practically infnite. Moreover, for
CMMs, evaluation tests can be performed to extract the
positioning errors, allowing correction models to be imple-
mented [1]. Tus, a high level of maintenance of the physical-
mathematical relations between the error model parameters
and the error physically committed by the machine can be
achieved. However, the application of these models does not
make sense in AACMMs, given the difculty of directly
Hindawi Publishing Corporation
e Scientic World Journal
Volume 2014, Article ID 681853, 11 pages
http://dx.doi.org/10.1155/2014/681853
2 Te Scientifc World Journal
relating the error committed with the model parameters,
which are obtained by using optimization procedures.
Te calibration procedure consists of identifying the geo-
metric parameters in order to improve the model accuracy.
Tis procedure allows us to obtain correction models to
establish corrections in the measurements results and to
quantify the efects of the infuence variables in the fnal
measurement. To achieve this goal, the following fve steps are
usually carried out: determination of the kinematic model by
means of nonlinear equations, data acquisition, optimization
or geometric parameter identifcation, model evaluation, and,
fnally, identifcation of the error sources and implementation
of correction models.
Te frst step, determination of the kinematic model, con-
sists of obtaining the non-linear equations that relate the joint
variables to the position and orientation of the end-efector
and the initial values of nominal geometric parameters. One
of the most widely used geometric methods for modelling a
mechanism is the well-known Denavit-Hartenberg method
[6], which models the joints with four parameters. One of the
limitations of this method appears in those mechanisms that
present two consecutive parallel joint axes. In this case, an
infnite number of common normals of the same length exist,
and the location of the axis coordinate systemmay be defned
arbitrarily. Some studies [710] present methods to obtain a
complete, equivalent, and proportional model.
Te number of parameters is fxed when the kinematic
model is determined, and this value will depend on the
selected method. Moreover, there are a maximum number of
parameters that must be identifed, and the model accuracy
does not have to improve by adding extra parameters [11].
In this work, the author determined that four parameters
must be considered for each revolute joint, two of which must
be orientational, and for a prismatic joint two orientational
parameters are necessary, applied about the noncollinear
axes before and perpendicular to the translational joint axes.
Nongeometric errors are usually compensated by adding
parameters in the geometric model [12].
Te second step is data acquisition. Any measurement
error of the external instrument is propagated to the results
of the identifed parameters. For that reason, an instrument
for data acquisition should be, at least, one magnitude order
more accurate than the mechanism whose parameters are
going to be identifed.
A direct geometric transformation can be established,
providing the global reference system can be measured by
means of an external measurement instrument such as a
laser tracker or a coordinate measuring machine. In this way,
this transformation obtains the coordinates of the measured
points in the global reference system of the mechanism.
Tus, direct comparisons in the objective function, between
nominal and measured data, can be made.
However, this relation is not usually easy to obtain
through a direct measure, and the problem is usually solved
by means of least-square methods. Tese methods allow us
to obtain an approximation of this transformation, and this
approximationwill depend onthe confgurations used indata
acquisition and on the mechanism error in the evaluated
points [13].
Te home position is a position, within the AACMM
working range, where all joint angles have a predefned
value. Te displacements of the probe are usually measured
with respect to this defned position. In [14], Kovac and
Frank developed a high precision gauge instrument for the
parameter identifcation procedure and evaluation tests.
Te determination of the number of required specifc
positions is not generalizable from one AACMM to another,
since the errors committed by each arm will depend on their
confguration and assembly defects. Te performance of tests
that characterize the infuence of each joint on the fnal error,
to fnally choose positions in accordance with this infuence,
must therefore be carried out before defning the capture
positions to identify the kinematic parameters. Te positions
selected for the identifcation procedure should cover the
maximum joint rotation range to cover the infuences of all
the measuring armelements in the workspace. In [15], Zheng
et al. obtained the spatial error distribution model by using
support vector machine theory.
Te third step is optimization or geometric parameter
identifcation, and the objective is to search for the optimum
values of all parameters included in the model that minimizes
the error in the performed measurements. Tis step is usually
carried out by means of approximation procedures based
on least-square ftting. Tis function can be defned as the
quadratic diference of the error (obtained between the mea-
sured value and the value computed by the kinematic model).
Te increment established for parameters must be defned for
each iteration. In most of the cases, numerical optimization
techniques are used to minimize the error. Te Levenberg-
Marquardt (L-M) [16] method is one of the most widely used
techniques to solve the numerical optimization algorithm.
Tis method usually presents lower computational cost,
providing a solution closer to the optimum solution for the
set of parameters considered. Moreover, the L-M algorithm
solves numerical problems that appear in other numerical
optimization techniques such as those based on the gradient
or on the least-square methods such as the Gauss-Newton.
A multiobjective optimization scheme is developed in [17] to
solve the nonlinear optimization problem. In [18], Santolaria
et al. presented a kinematic parameter estimation technique,
which allows us to improve the repeatability of the AACMM
by more than 50%. Tis technique uses a ball bar gauge to
perform the data acquisition procedure.
Te fourth step, the model evaluation, consists of evalu-
ating the mechanism behavior with the set of the identifed
parameters obtained in the geometric parameter identifca-
tion procedure, in confgurations diferent from those used
in the optimization process. In [19], Koseki et al. evaluated
the accuracy of a mechanism by means of a laser tracking
coordinate measuring system.
Finally, an identifcation of the error sources and a
modelling and implementation of the correction models can
be optionally performed.
In [20], Piratelli-Filho et al. developed a virtual sphere
plate, having a standard deviation of around 0.02 mm and
a measurement uncertainty from 0.02 to 1.64 m in point
measurements, to evaluate the measurement performance of
AACMMs. Te AACMM performance test using the virtual
Te Scientifc World Journal 3
sphere plate resulted in a mean error of 0.023 mm and a
standard deviation of 0.039 mm.
Alaser tracker is a large-scale measuring instrument with
highaccuracy. Tese mechanisms are consideredvery reliable
[21]. Consequently, laser trackers have been used recently
instead of other traditional methods such as theodolites or
collimators in multiple applications such as robot tracking,
testing, calibration, and maintenance. Tese systems use
interferometry for measuring relative distances and optical
encoders in order to measure azimuth and elevation angles
of a beam-steering mirror. Te interferometer measurements
are obtained relative to the starting point. Besides, this beam
must track the positions of a retrorefector. A plane mirror
mounted on a high precision universal joint defects the
beam and hits the retrorefector. Tis element consists of
three perpendicularly oriented plane mirrors, and the beamis
refected parallel. Teoretically, the laser beam hits the center
point of the retrorefector. When there is no relative move-
ment between the laser tracker head and the retrorefector,
there is no parallel displacement between the emitted and
the refected beam. However, when the retrorefector starts
moving, there is a displacement of the refected laser beam,
since, in this case, the laser beam does not hit the center
point of the retrorefector. In [22], Lin and Her used a laser
tracker systemto measure the volumetric errors of a precision
machine. Te technique presented is based on the ASME
B5.54 standards and ofers a quick method to characterize
these errors. Tis measuring instrument can also be used in
multilateration for verifcation of machine tools, as presented
in [23], where the error order can also be in the error range of
a laser tracker. In [21], a laser tracker was used also to analyse
the performance of an indoor GPS.
Te multilaterationtechnique is a procedure whichcanbe
usedtoimprove the mechanismaccuracy. Tis methodallows
us to reduce the measurement uncertainty by eliminating the
angular noise. To achieve this, the multilateration obtains
the half point position starting from data of several laser
trackers, located in diferent positions [24]. In [25], Zhang
et al. presented specifc recommendations for optimization
of multilateration set-ups and measurement plans and for
minimizing measurement uncertainty. Besides, the authors
evaluated the volumetric measurement error propagation,
obtaining an average standard deviation of length mea-
surements around 1.2 m. Hughes et al. [26] presented a
laser-interferometric measuring station and obtained the
displacement measurement uncertainty, which was used to
predict a volumetric uncertainty for a multilateration system,
consisting of eight measuring stations and four targets. Te
combined uncertainty of the measuring station displacement
measurements and the CMM repeatability obtained was
around 200 nm. In [27], Kim et al. developed a volumetric
interferometer system and minimized least-square errors
by ftting the measured values to a geometric model of
multilateration, obtaining a volumetric uncertainty of less
than 1 m.
Te aim of this work is to improve the accuracy of an
AACMM by means of a new calibration procedure based
on laser tracker multilateration. Although both instruments
present the same order of magnitude with respect to the
accuracy, the multilateration techniques allow us to capture
points with an uncertainty much smaller than the one
obtained with an AACMM. Te points captured from the
multilateration are therefore considered as nominal data in
the calibration procedure of the AACMM.
2. AACMM Kinematic Model
Te AACMM kinematics relates the joint variables and the
probe position for any arm posture.
Te direct kinematic model is used to calculate the
positioning and orientation of the AACMM probe on the
basis of certain values of the joint variables, according to the
following:
= (

, ) , (1)
with = 1, . . . , for an arm with rotating joints.

is given
by the vector of the joint variables and represents the model
defned, which depends on the parameter vector .
Te non-linear equation system to model the mechanism
can be developed by applying the Denavit-Hartenberg (D-H)
method[6] toevery chaininthe mechanism. Tis methodhas
been widely used in mechanism modelling [28, 29] and uses
four parameters (distances

, and angles

) to model
the coordinate transformation between successive reference
systems. Te homogenous transformation matrix between
frame and 1 depends on these four parameters:
1

=
,

,

,

,
=
[
[
[
[
[
cos

cos

sin

sin

sin

cos

sin

cos

cos

sin

cos

sin

0 sin

cos

0 0 0 1
]
]
]
]
]
.
(2)
Te AACMM model used in this work is a Faro Platinum
Arm having seven axes, with a nominal value of 2 in the
single point articulation and performance test of 0.030 mm
according to the specifcations of the manufacturer.
Figure 1 shows the reference systems usedinthe AACMM
model.
Te armglobal transformationmatrix allows us toexpress
the probe sphere center coordinates with respect to the base
of the AACMM. Tis matrix can be obtained by calculating
successive coordinate transformations by premultiplying the
transformation matrix between a frame and the previous one,
as shown in the following:
[, , , 1]

AACMM
=
0

6
[, , , 1]

Probe
. (3)
In this equation, 0 is the global reference system of the
base and 6 corresponds to the reference frame that moves
with the rotation of the last joint.
A reference system is usually defned in the probe.
However, the aim of this study is to obtain the sphere center
4 Te Scientifc World Journal
X
0
Y
0
Z
0
X
4
Y
4
Z
4
X
5
Y
5
Z
5
X
6
Y
6
Z
6
X
7
Y
7
Z
7
X
3
Y
3 Z
3
X
2
Y
2
Z
2
X
1
Y
1
Z
1
Figure 1: AACMM reference systems.
coordinates of the static probe, so this reference system is
not necessary. Seven reference systems are used to model the
AACMM. Te last reference system is located in the center
of the refector and is oriented as reference system six. Tus,
the number of parameters for the 6 degrees of freedom (Dof)
AACMM is 28. Te kinematic model is fully described in
[18]. Table 1 shows the initial values for the AACMM D-H
parameters.
3. Multilateration System
Te multilateration technique allows us to reduce the mea-
surement uncertainty by eliminating the angular noise. To
achieve this, the multilateration obtains the weighted-point
position by means of several laser tracker data, located in
diferent positions. A minimum of three measurements of
each point is necessary. Each laser tracker measures the
distance from the laser tracker to the target points. Tese
measurements present a noise having a radial component and
two angular components. Te aim of the multilateration is to
decrease the measurement uncertainty, so this technique only
uses the radial component of the laser tracker measurements,

, thus decreasing the measurement noise infuence, which


allows us to decrease the global uncertainty. Tese compo-
nents defne a sphere. Te intersection of the three spheres
obtained by measuring the same point by three laser trackers
provides two points. Equations from (4) to (6) provide the
spheres obtained by the measurements of each laser tracker,
respectively. Consider

2
0
= (
1
)
2
+ (
1
)
2
+ (
1
)
2
, (4)

2
1
= (
2
)
2
+ (
2
)
2
+ (
2
)
2
, (5)

2
2
= (
3
)
2
+ (
3
)
2
+ (
3
)
2
. (6)
Te knowledge of the locations of the reference systems of
the laser trackers allows us to obtain the following equations:
=
(
2
0

2
1
+
2
2
)
2
1
,
(7)
Reector (X, Y, Z)
m
3
P
3
(X
3
, Y
3
, Z
3
)
m
0
m
2
m
1
P
2
(X
2
, Y
2
, 0)
P
0
(0, 0, 0)
P
1
(X
1
, 0, 0)
Figure 2: Coordinates of the four laser tracker reference systems
expressed in the multilateration system.
=
(
2
0

2
2
+
2
2
+
2
2
2
2
)
2
2
,
(8)
= (
2
0

2

2
)
1/2
.
(9)
Inthis work, for simplicity, the unknownreference system
location of the three laser trackers has been defned as
(0, 0, 0), (
1
, 0, 0), and (
2
,
2
, 0) with respect to the multilat-
eration reference system.
A fourth laser tracker can be used to avoid the sign
ambiguity in the coordinate obtained in (9). Te reference
system of this laser tracker is given by (
3
,
3
,
3
). Tis point
should belong to a plane diferent from the plane formed
by the other three laser tracker reference systems as shown in
Figure 2. In this fgure,

represents LT

position system for


the laser trackers. Te origin of the multilateration global
reference system is given by
0
.
Te (, , ) target coordinates are obtained in linear
matrix form by operating (7), (8), and (9), as expressed in the
following:
[
[

]
]
= 0,5
[
[
[
[
[
[
[
[
[
[
[
1

1
0 0

2
1

2
0
(

3

3
) + (

2

3
)

3

3
1

3
]
]
]
]
]
]
]
]
]
]
]

[
[
[
[
[

2
1

2
0

2
1

2
2

2
0

2
2

2
2

2
3

2
0

2
3

2
3

2
3
]
]
]
]
]
.
(10)
Te multilateralized coordinates, obtained in (10), will be
considered as the nominal coordinates in the identifcation
parameter procedure.
4. Data Acquisition
Te data acquisition step consists of capturing the nominal
coordinates in the workspace of the AACMM. Te suitable
number of positions is not generalizable from one measuring
Te Scientifc World Journal 5
Table 1: Initial values for the AACMM D-H parameters.
Joint

(mm)

(mm)
1 90 90 40 300
2 135 90 40 0
3 180 90 34 590
4 90 90 34 0
5 90 90 34 588
6 180 90 34 0
7 0 0 0 318.2
LT
4
: FARO ION
LT
2
: API T3
LT
1
: FARO X
LT
3
: LEICA LT-600
Retroreector
AACMM
Figure 3: LTs and AACMM distribution used in tests.
arm to another, since each measuring arm error will depend
on their confguration and assembly defects. Te identifca-
tion procedure should cover the maximum range of joint
rotation to consider all the infuences of the measuring arm
elements.
In this study, a cloud of points located within the
arm workspace was measured simultaneously with both the
AACMM (measured values) and four LTs which conform
to the multilateralized system (nominal values). Te laser
tracker FARO model was used as LT
1
, API T3 as LT
2
,
LEICA LT-600 as LT
3
, and FARO ION as LT
4
. Te positions
were distributed throughout the workspace of the arm and
reached diferent arm angle values. Te LTs were distributed
forming the multilateration global coordinate system, as it
was detailed in Section 3. LT

positions have been chosen in


function of the refector visibility, thus forming a spatial angle
as near as possible to 90

between them. Te measurement


uncertainty is lower in this position, as demonstrated in a
previous work [30]. Te AACMM has been arranged in a
position that maximizes the visibility of the LTs, as shown in
Figure 3.
Te AACMM data acquisition technique is usually per-
formed by means of discrete contact probing of surface points
of the gauge in order to obtain the center of the spheres
from several surface measurements. Te time required for
the capture of positions is high, and therefore, identifcation
120
X
center
, Y
center
, Z
center
L
P
r
o
b
e
45

120

Figure 4: Probe used in the data acquisition procedure.


is generally carried out with a relatively low number of arm
positions. In this work, a probe presented in [18], capable
of directly probing the center of the spheres of the gauge
without having to probe surface points, has been used. Tis
probe consists of three tungsten carbide spheres of 6 mm in
diameter, laid out at 120

on the probe, as can be seen in


Figure 4.
Te probe usedallows us to defne a probe withzero probe
sphere radius and with a distance from the position of the
housing to the center of the probed refector sphere of 1.5
inches, allowing direct probing of the sphere center when the
three spheres of the probe and the sphere are in contact.
One of the advantages of this type of probe is that
the massive capture of arm positions can be performed
corresponding to several points of the workspace, whichleads
to save a considerably amount of time.
23 positions of the retrorefector were measured, from
which 21 positions were considered in the parameter iden-
tifcation process (identifcation positions) and the other 2
positions were kept for the parameter evaluation procedure
(test positions). A sofware developed captured the AACMM
measurements, saving the AACMM joint angles,

. Tese
angles and the AACMM parameters are the input to the
kinematic model. Te solution of the non-linear equations by
the L-Malgorithmobtains the measured point coordinates in
the AACMM reference system.
Te data acquisition procedure was performed trying to
capture data in symmetrical trajectories in the retrorefector
to minimize the efect of probing force on the gauge.
Although the measuring of the retrorefector center with
the kinematic mount probe from diferent arm orientations
should result in the same point measured, the unsuitable
value of the nominal kinematic parameters of the model will
be shown by way of a probing error, resulting in diferent
coordinates for the same measured point in diferent arm
orientations. For that reason, fve measurements were taken
for eachpoint and the meanpoint of the set of points captured
was considered as the center of the retrorefector measured in
the AACMM reference system, as shown in Figure 5.
Te distance between the diferent retrorefector mea-
sured positions can be obtained from the points captured
by means of the Euclidean distance between each pair of
retrorefector positions, obtaining
AACMM
, where and
represent two measured positions.
At the same time, the four LTs simultaneously measured
the distance from the captured point to the local coordinate
system.
6 Te Scientifc World Journal
X
c
, Y
c
, Z
c
=
X
i
, Y
i
, Z
i
X
i
, Y
i
, Z
i
Figure 5: Retrorefector center.
Diagonal distances are obtained according to the follow-
ing:
= + , (11)
where is the incremental output of the displacement
transducer used and is the ofset, which should be calibrated
before the multilateration implementation.
Te self-calibration and determination of the parameter
can be carried out by capturing some additional objective
measurements in some diferent positions. By performing
the quadrilateration objective point times, the number of
equations is given by 4 as shown in the following:
(

)
2
= (

)
2
= (

)
2
+ (

)
2
+ (

)
2
,
(12)
for = 0, . . . , 3 and = 0, . . . , 1.
Te system unknowns are the four ofsets,

, and the
position of the coordinates,

(3 unknowns). To
identify the four references, 12 unknowns, corresponding
to

, must be added. Tus, the number of system


unknowns is given by 3 + 16.
Te unknowns can be obtained by the least-square
method to minimize the error.
Te multilateration technique must solve the objective
function defned by
=
21

=1
4

=1
[(


LT

)
2
+ (


LT

)
2
+ (


LT

)
2
(

+
LT

)
2
] ,
(13)
where (

) represents the measured point coordinates


in the multilateration reference system, (
LT

,
LT

,
LT

) is
the origin of the LT

reference system,

are the measured


distances for every point in the LT

reference system, and


LT

represents the LT

ofset. Te non-linear system obtained by


the multilateration technique was solved by means of the L-
M algorithm, and the solution obtained gives the following
information:
(i) the point coordinates in the multilateration system:
(

)|
SR Multi
,
(ii) the laser tracker ofsets:
LT

,
(iii) the origin of the reference systems for the four laser
trackers in the multilateration system: (
LT

,
LT

LT

)|
SR Multi
.
Te calculated distances between each pair of retrorefec-
tor positions can be obtained from the point coordinates in
the multilateration system, obtaining
Multi
, where and
represent the two positions.
5. Parameter Identification Procedure
Once the data acquisition technique has been carried out, the
parameter identifcation procedure can be performed.
Figure 6 shows a scheme of the calibration procedure.
Te origin of the laser trackers was measured with respect
to the reference systemof LT
1
. Tis reference systemhas been
considered the multilateration reference systemfor obtaining
the initial values of the LT

reference systems.
Te kinematic parameter identifcation procedure can be
performed starting from the measured and calculated dis-
tances (obtained as explained in Section 4). Te parameters
considered in the AACMM kinematic model are given by
(

,
0
,
Enc
), where
Enc
are the angles measured by
the encoder. Te model obtained is a non-linear equation
system, and it was solved by the L-Malgorithm. Te objective
function defned in this step considers both the measured and
calculated distances, as shown in
=

=1,=1
[(

AACMM

Multi
)
2
] +

=1
[
2

+
2

+
2

] ,
(14)
where represents the number of positions considered
in the parameter identifcation procedure and (

) represents the standard deviation of the points mea-


sured in each position and each coordinate, showing the
infuence of the volumetric accuracy and point repeatability.
6. Calibration Results
In the optimization process, the distances from each point to
every point are taken into account, obtaining 253 distances
between the 23 points.
Te multilateration procedure described above was car-
ried out based on the 23 captured points for each LT
corresponding to the refector positions used. Table 2 shows
the results obtained. As the initial values for this procedure,
null values were assigned for the ofsets of all LTs, and for the
origin points the corresponding coordinates were those that
better ft the distribution observed taken from the measured
home points of LT
2
, LT
3
, and LT
4
fromLT
1
, where the origin
of the multilaterized reference system is located.
In Figure 7, the coordinates of the captured points by
each LT and the coordinates of the multilaterated points,
Te Scientifc World Journal 7
Measuring each point with the
AACMM coordinates
Direct model
(D-H)
End
End
Home
Initial values
Multilateration
Measuring each point with all the LT
AACMM coordinates
Parameter identifcation
m
ij
(x
LTj
, y
LTj
, z
LTj
)
SR LT1
(x
LTj
, y
LTj
, z
LTj
)
SR multil.
I
LTj
I
LTj
initial
AACMM:
i
(21 positions, 5 measurements
for each position)
(x
i
, y
i
, z
i
)
SR multil.
(x
i
, y
i
, z
i
)
SR AACMM
d
ij multil. d
ij AACMM
For iter = 1 to (F
error
< tolerance)
For i, j = pos ini to pos fnal
=
21

i=1,j=1
[(d
ij
AACMM
d
ij
multil
)
2
+
n

i=1
[
2
x

+
2
y

+
2
z

]
Optimisation: to minimize F
error
Calculation of the geometric parameters: S
i
values
=
21

i=1
4

j=1
[(x
i
x
LTj
)
2
+ (y
i
y
LTj
)
2
+ (z
i
z
LTj
)
2
(m +
ij
l
LTj
)
2
]
F
error
= [E
1
. . . E
pos final
]
Figure 6: Scheme of the calibration procedure.
used as nominal coordinates in the parameter identifcation
procedure of the AACMM, are shown graphically.
As stated before, in order to introduce redundancy in
the objective function and thereby restrict to the nominal
points the fnal points obtainedwiththe identifedparameters
of the measuring arm, all possible distances between the
multilaterated points are calculated. Tis way, we can obtain
253 distances, which will be used as nominal data in the
objective function (14) in the identifcation procedure. In
Figure 8, the range of each of the calculated distances in the
measurements made by the four LTs can be observed. Tus,
each data represents the diference between the maximum
and minimum values of each considered distance calculated
from the set of the 4 LTs measured data. Te distances are
arranged starting from position 1 of the retrorefector. Te
maximum range value obtained by calculating the distances
with the 4 LTs was 105 m, while the mean range for all set of
distances was 43 m.
Following the scheme presented in Figure 6, we can
calculate all possible distances between spheres as well as
the standard deviations measured with the AACMM from
the saved angular data obtained during the data capture
process and the set of initial parameters of the AACMM
mathematical model shown in Table 1. In accordance with
the optimization for the identifcation scheme presented in
Figure 6, the quality indicators for the set of initial parameters
of the AACMM model are shown in Table 3. Moreover,
Table 3 shows inits frst columnthe distances maximumerror
obtained for all the refector points and the index of the points
that determine the distance in which the maximum error is
calculated.
Analogously, the mean distance errors for all the evalu-
ated distances are also shown. With respect to the standard
deviation, Table 3 shows, besides the maximum and mean
values, the index of the point and its corresponding coordi-
nate where the greater value is obtained, since in this case the
objective function considers the standard deviation for each
coordinate independently. As expected, the obtained values
are high, considering the initial set of parameters defned for
the AACMM mathematical model.
In the objective function proposed in (14), the data
capture setup described for = 21 refector points, it is
necessary to consider the elimination of the terms where =
to avoid both the inclusion of null terms and duplicate of
8 Te Scientifc World Journal
Table 2: Multilaterated ofsets and origin coordinates obtained in multilateration reference system. 2798 iterations and objective function
value below 1 m.
Ofsets (mm)
Origin coordinates (mm)

LT

LT

LT
LT1
LT1
1.51695 0 0 0
LT2
LT2
0.89275 2262.33138 0 0
LT3
LT3
0.18749 1713.00335 2098.93521 0
LT4
LT4
2.13351 1913.00558 343.41959 3010.23578
Table 3: Quality indicators for the initial values of model parameters over 21 SMR locations (105 AACMM positions).
Distance error (mm) 2 by retrorefector point (mm)
Maximum 38.50723 Maximum 22.00413
Causing dist. 410 Causing point 17
Medium 17.28859 Causing coord. X
Medium 8.17533
Table 4: Identifed values for the AACMM model parameters by L-
M algorithm.
Joint

(mm)

(mm)
1 89.1269 89.9216 42.4313 300
2 134.7734 89.7178 41.7885 1.2337
3 184.2704 90.1202 28.9084 591.2423
4 95.4182 89.5364 29.4449 0.6300
5 88.0773 89.9482 28.5745 591.5284
6 181.0832 92.5750 27.5479 12.2143
7 0.2117 0 0.4124 254.0575
distance errors noting that

. In regard to the standard


deviation, each coordinate deviation for every one of the
refector positions is considered. To express mathematically
the optimization problem it is required to consider the sum
of the calculated quadratic errors. Tis way, through the
objective function in (14), we obtain 231 terms corresponding
to the distance errors for the 21 refector points, plus 63
standard deviation terms of these points, obtaining a total
of 294 terms to determine the objective function value in
each iteration of the optimization algorithm. Tis value will
contain the infuence of the kinematic parameters and the
articulation variables considering for this the terms related
on the one hand to volumetric precision and on the other to
repeatability, for the 105 measuring arm captured positions.
In Figure 9, it can be observed graphically the distribution of
the captured points in the AACMM reference system, before
and afer the optimization procedure, while Table 4 shows
the identifed parameters starting from the initial values of
Table 1.
From the identifed parameters (Table 4), Table 5 shows
the error characteristics results obtained for each of the
captured points considering in this case these parameters.
Te validation and generalization of the error results cal-
culated for the set of identifed parameters over the captured
data, to the rest of the AACMM work volume, imply in the
more restrictive case obtaining error and deviation values less
than the maximum values obtained in this case (Table 5) for
any evaluated position of the measuring arm. For this reason,
the assessment of the AACMM error in diferent positions
to the ones used in the identifcation procedure is highly
recommended. As shown in Table 5, a maximum error of
118 m and a mean error of 48 m for the measuring volume
have been obtained, considering the 21 nominal points used
in the identifcation. In normal operation of the measuring
arm in this work volume, it is expected to get error values
close to the mean value and obtaining maximumerror values
only in certain arm confgurations.
As a last step of the identifcation procedure, it is
necessary to evaluate the set of parameters obtained in
diferent arm positions from those considered in its own
identifcation procedure, such that it is possible to conclude
that the error results can be considered reliably within the
measuring arm work volume. Furthermore, it is expected
that the more similar the measuring arm positions are,
when probing the refector points, to the ones used in the
identifcation, the closer the error results should be to the
ones shown in Table 5. Terefore, the points and positions
for evaluation must be diferent from the ones used in the
identifcation procedure. To illustrate this characteristic, in
this case two extra positions of the refector have been
considered as test points (Figure 7). For each one of these
new refector positions, 10 angle combinations have been
captured corresponding to the center positions of each one
of them, captured in the same captured conditions compare
to the rest of the points. Te nominal distance between
these two points, calculated as the Euclidean distance of the
multilaterated coordinates, was 492.7164 mm. Te distance
obtained, using the mean points expressed in the AACMM
reference systemof the 10 probed points of each point and the
identifed parameters, was 492.7695 mm, obtaining an error
in this case with respect to the nominal of 53 m, while the
maximum standard deviation for the two probed points has
been calculated in the coordinate of the frst probed point,
with a value of 0.1426 mm for all the 10 captured positions.
It is therefore possible to conclude that the obtained error
value for the identifed parameter set can be generalizable to
Te Scientifc World Journal 9
Table 5: Quality indicators for the identifed values of model parameters over 21 SMR locations (105 AACMM positions).
Distance error (mm) 2 by retrorefector point (mm)
Maximum 0.11824 Maximum 0.20644
Causing dist. 212 Causing point 4
Medium 0.04825 Causing coord. Y
Medium 0.11545
500
1000
1500
3000
3500
2000
1800
1600
1400
1200
Identifcation points
Evaluation points
X
(
m
m
)
Z
(
m
m
)
Y (mm)
(a)
2000
0
2000
4000
2000
4000
1000
500
0
500
1000
X
(
m
m
)
Z
(
m
m
)
LT
1
LT
2
LT
3
LT
4
Y
(m
m
)
(b)
Figure 7: Points captured by LTs. (a) Multilaterated points used as nominal data for the parameter identifcation procedure; (b) points
captured by each LT.
0 50 100 150 200 250
0
20
40
60
80
100
120
Distance number
R
a
n
g
e

(

m
)
Figure 8: Range of each distance considered in the identifcation
process. Each data represents the diference between the maximum
and minimum values of each distance calculated with the measure-
ments made by the four LTs.
the evaluated work volume. For this reason, the assessment
of more evaluation positions diferent from the one used
as an example is recommended when increasing the work
volume to be identifed. Te ideal is to obtain error values
always below the identifcation maximum error, although
it is possible to set an acceptable error percentage above
the obtained maximum error in order to set a characteristic
value of the measuring arm global error according to a less
restrictive criterion.
7. Conclusions
In this work, a novel calibration technique for parameter
kinematic identifcation of an AACMM is presented. Tis
calibration technique is based on an objective function that
considers the volumetric error and repeatability by means
of the distance errors and the standard deviation of physical
probed points, respectively. Moreover, a new procedure to
obtain nominal gauge values for this calibration technique
is carried out. Tis new procedure is based on the mea-
surements of a calibrated spherical retrorefector with 4 LTs.
Although the error range of this type of measuring instru-
ments has an order of magnitude similar to the AACMM
for this work distances, the use of multilateration techniques
can be of great help to reduce the measurement uncertainty,
taking as nominal data only the measurements of the LTs.
By combining the aforementioned measurements and afer
10 Te Scientifc World Journal
1000
0
1000
1000
0
1000
1500
1000
500
0
500
1000
1500
X
(m
m
)
Z
(
m
m
)
Y
(m
m
)
(a)
X
(
m
m
)
Z
(
m
m
)
200
400
600
800
1000
200
0
200
400
600
800
150
200
250
300
350
400
450
Y
(m
m
)
(b)
Figure 9: Points captured by the AACMM. (a) Nominal kinematic parameters; (b) identifed kinematic parameters.
the described optimization procedure, is possible to obtain
points that can be used as nominal points, in this case
materializing distances between them for their use in the
AACMMparameter identifcationprocedure. Eventhoughin
this work 4 LTs have been used to eliminate the coordinate
sign ambiguity of the multilaterated points, it is possible
to realize this procedure using only 3 LTs by making sure
that all of the points have the same sign with respect to the
multilateration reference system. Tis way, in the cases when
access to this type of measuring instruments is available, it
is possible to carry out an AACMM identifcation procedure
without the use of common physical gauges used in this type
of procedures.
Finally, the simplifcation of the calibration procedure
presented in this work can be achieved by applying sequential
multilarization; thus the use of only one LT is needed to carry
out the adapted procedure, with the aim of reducing costs.
Conflict of Interests
Te authors declare that there is no confict of interests
regarding the publication of this paper.
References
[1] A. Piratelli-Filho and G. R. Lesnau, Virtual spheres gauge for
coordinate measuring arms performance test, Measurement,
vol. 43, no. 2, pp. 236244, 2010.
[2] K. Shimojima, R. Furutani, K. Takamasu, and K. Araki, Te
estimation method of uncertainty of articulated coordinate
measuring machine, in Proceedings of the IEEE International
Conference on Industrial Technology, pp. 411415, Dubrovnik,
Croatia, 2002.
[3] H. Hamana, M. Tominaga, M. Ozaki, and R. Furutani,
Calibration of articulated arm coordinate measuring machine
considering measuring posture, International Journal of
Automation Technology, vol. 5, no. 2, pp. 109114, 2011.
[4] J. Sadek, K. Ostrowska, and A. Gska, Modeling and identif-
cation of errors of coordinate measuring arms with the use of a
metrological model, Measurement, vol. 46, no. 1, pp. 667679,
2013.
[5] J. Sadek, K. Ostrowska, and K. Gacek, Kinematic metrological
model of the coordinate measuring arm (MCMA), in Pro-
ceedings of the 19th IMEKO World Congress Fundamental and
Applied Metrology, pp. 19881992, Lisbon, Portugal, 2009.
[6] J. Denavit and R. S. Hartenberg, A kinematic notation for
lower-pair mechanisms based on matrices, Transactions of
ASME-Journal of Applied Mechanics, vol. 22, no. 1, pp. 215221,
1955.
[7] M. Miroir, Y. Nguyen, J. Szewczyk, O. Sterkers, and A. Bozorg
Grayeli, Design, kinematic optimization, and evaluation of a
teleoperated system for middle ear microsurgery, Te Scientifc
World Journal, vol. 2012, Article ID 907372, 19 pages, 2012.
[8] T. W. Hsu and L. J. Everett, Identifcation of the kinematic
parameters of a robot manipulator for positional accuracy
improvement, in Proceedings of the Computation in Engineering
Conference, pp. 263267, Boston, Mass, USA, August 1985.
[9] B. W. Mooring, Te efect of joint axis misalignment on
robot positioning accuracy, in Proceedings of the Computers in
engineering: International Computers in Engineering Conference
and Exhibit, pp. 151155, Chicago, Ill, USA, August 1983.
[10] H. Stone, A. Sanderson, and C. Neuman, Armsignature identi-
fcation, in Proceedings of the IEEE International Conference on
Robotics and Automation, pp. 4148, San Francisco, Calif, USA,
April 1986.
[11] L. J. Everett and A. H. Suryohadiprojo, A study of kinematic
models for forward calibration of manipulators, in Proceedings
of the IEEE International Conference on Robotics and Automa-
tion, pp. 798800, Philadelphia, Pa, USA, April 1988.
Te Scientifc World Journal 11
[12] J. Chen and L. M. Chao, Positioning error analysis for robot
manipulators with all rotary joints, IEEE Journal of Robotics
and Automation, vol. RA-3, no. 6, pp. 539545, 1987.
[13] A. C. Majarena, J. Santolaria, D. Samper, and J. J. Aguilar,
An overview of kinematic and calibration models using inter-
nal/external sensors or constraints to improve the behavior of
spatial parallel mechanisms, Sensors, vol. 10, no. 11, pp. 10256
10297, 2010.
[14] I. Kova c and A. Frank, Testing and calibration of coordinate
measuring arms, Precision Engineering, vol. 25, no. 2, pp. 90
99, 2001.
[15] D. Zheng, C. Du, and Y. Hu, Researchonoptimal measurement
area of fexible coordinate measuring machines, Measurement,
vol. 45, no. 3, pp. 250254, 2012.
[16] K. Levenberg, A method for the solution of certain non-linear
problems in least squares, Quarterly of Applied Mathematics,
vol. 2, no. 2, pp. 164168, 1944.
[17] Q. Gao, J. Chen, L. Wang, S. Xu, and Y. Hou, Multiobjective
optimization design of a fractional order PID controller for
a gun control system, Te Scientifc World Journal, vol. 2013,
Article ID 907256, 8 pages, 2013.
[18] J. Santolaria, J. Aguilar, J. Yag ue, and J. Pastor, Kinematic
parameter estimation technique for calibration and repeata-
bility improvement of articulated arm coordinate measuring
machines, Precision Engineering, vol. 32, no. 4, pp. 251268,
2008.
[19] Y. Koseki, T. Arai, K. Sugimoto, T. Takatuji, and M. Goto,
Design and accuracy evaluation of high-speed and high
precision parallel mechanism, in Proceedings of the 1998 IEEE
International Conference on Robotics and Automation. Part 1 (of
4), pp. 13401345, May 1998.
[20] A. Piratelli-Filho, F. H. T. Fernandes, and R. V. Arencibia,
Application of virtual spheres plate for AACMMs evaluation,
Precision Engineering, vol. 36, no. 2, pp. 349355, 2012.
[21] Z. Wang, L. Mastrogiacomo, F. Franceschini, and P. Maropou-
los, Experimental comparison of dynamic tracking perfor-
mance of iGPS and laser tracker, International Journal of
Advanced Manufacturing Technology, vol. 56, no. 14, pp. 205
213, 2011.
[22] C. Lin and J. Her, Calibrating the volumetric errors of a
precision machine by a laser tracker system, International
Journal of Advanced Manufacturing Technology, vol. 26, no. 11-
12, pp. 12551267, 2005.
[23] S. Aguado, D. Samper, J. Santolaria, and J. J. Aguilar, Identif-
cation strategy of error parameter in volumetric error compen-
sation of machine tool based on laser tracker measurements,
International Journal of Machine Tools and Manufacture, vol. 53,
no. 1, pp. 160169, 2012.
[24] T. Takatsuji, M. Goto, A. Kirita, T. Kurosawa, and Y. Tanimura,
Te relationship between the measurement error and the
arrangement of laser trackers in laser trilateration, Measure-
ment Science and Technology, vol. 11, no. 5, pp. 477483, 2000.
[25] D. F. Zhang, S. Rolt, and P. G. Maropoulos, Modelling and
optimization of novel laser multilateration schemes for high-
precision applications, Measurement Science and Technology,
vol. 16, no. 12, pp. 25412547, 2005.
[26] E. B. Hughes, A. Wilson, and G. N. Peggs, Design of a high-
accuracy CMM based on multi-lateration techniques, CIRP
Annals-Manufacturing Technology, vol. 49, no. 1, pp. 391394,
2000.
[27] S. W. Kim, H. G. Rhee, and J. Y. Chu, Volumetric phase-
measuring interferometer for three-dimensional coordinate
metrology, Precision Engineering, vol. 27, no. 2, pp. 205215,
2003.
[28] J. Santolaria, A. Brau, J. Velzquez, and J. J. Aguilar, A self-
centering active probing technique for kinematic parameter
identifcation and verifcation of articulated arm coordinate
measuring machines, Measurement Science and Technology,
vol. 21, no. 5, Article ID 055101, 2010.
[29] W. K. Kim, Y. K. Byun, and H. S. Cho, Closed-form forward-
position solution for a 6-DoF 3-PPSP parallel mechanism and
its implementation, International Journal of Robotics Research,
vol. 20, no. 1, pp. 8599, 2001.
[30] S. Aguado, J. Santolaria, D. Samper, and J. Aguilar, Infuence
of measurement noise and laser arrangement on measurement
uncertainty of laser tracker multilateration in machine tool
volumetric verifcation, Precision Engineering, vol. 37, pp. 929
943, 2013.
Submit your manuscripts at
http://www.hindawi.com
VLSI Design
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
International Journal of
Rotating
Machinery
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com
Journal of
Engineering
Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Shock and Vibration
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Mechanical
Engineering
Advances in
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Civil Engineering
Advances in
Acoustics and Vibration
Advances in
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Electrical and Computer
Engineering
Journal of
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Distributed
Sensor Networks
International Journal of
The Scientifc
World Journal
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Sensors
Journal of
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Modelling &
Simulation
in Engineering
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Active and Passive
Electronic Components
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Chemical Engineering
International Journal of
Control Science
and Engineering
Journal of
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Antennas and
Propagation
International Journal of
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Navigation and
Observation
International Journal of
Advances in
OptoElectronics
Hindawi Publishing Corporation
http://www.hindawi.com
Volume 2014
Robotics
Journal of
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014